linux-next build conflict between modules and metag trees (LOCKDEP_NOW_UNRELIABLE)

linux-next build conflict between modules and metag trees (LOCKDEP_NOW_UNRELIABLE)

[James Hogan]

Hi Rusty,

The metag architecture tree adds an add_taint(TAINT_DIE) like other
architectures do, and the modules-next tree adds the
LOCKDEP_NOW_UNRELIABLE flag to all uses of add_taint (but obviously
misses arch/metag since it doesn't exist yet), causing a compile error
on metag in -next when the two are merged together.

Is it okay for me to merge your commit 373d4d0 ("taint: add explicit
flag to show whether lock dep is still OK.") in modules-next into the
base of the metag tree and expect it not to be rebased, so that I can
then squash the fix into the metag tree?

The only commits this would include are:
$ git log --oneline linus/master..373d4d0
373d4d0 taint: add explicit flag to show whether lock dep is still OK.
64748a2 module: printk message when module signature fail taints kernel.

Thanks
James

Re: linux-next build conflict between modules and metag trees (LOCKDEP_NOW_UNRELIABLE)

[Rusty Russell]

James Hogan <james.hogan@imgtec.com> writes:
> Hi Rusty,
>
> The metag architecture tree adds an add_taint(TAINT_DIE) like other
> architectures do, and the modules-next tree adds the
> LOCKDEP_NOW_UNRELIABLE flag to all uses of add_taint (but obviously
> misses arch/metag since it doesn't exist yet), causing a compile error
> on metag in -next when the two are merged together.
>
> Is it okay for me to merge your commit 373d4d0 ("taint: add explicit
> flag to show whether lock dep is still OK.") in modules-next into the
> base of the metag tree and expect it not to be rebased, so that I can
> then squash the fix into the metag tree?

This was my fault for taking a shortcut.  I should have changed the name
so the old add_taint worked still (set_taint?), then remove add_taint
after the merge.

But I won't be rebasing, so you should be fine to merge it.

Cheers,
Rusty.

Re: linux-next build conflict between modules and metag trees (LOCKDEP_NOW_UNRELIABLE)

[James Hogan]

On 08/02/13 03:17, Rusty Russell wrote:
> James Hogan <james.hogan@imgtec.com> writes:
>> Hi Rusty,
>>
>> The metag architecture tree adds an add_taint(TAINT_DIE) like other
>> architectures do, and the modules-next tree adds the
>> LOCKDEP_NOW_UNRELIABLE flag to all uses of add_taint (but obviously
>> misses arch/metag since it doesn't exist yet), causing a compile error
>> on metag in -next when the two are merged together.
>>
>> Is it okay for me to merge your commit 373d4d0 ("taint: add explicit
>> flag to show whether lock dep is still OK.") in modules-next into the
>> base of the metag tree and expect it not to be rebased, so that I can
>> then squash the fix into the metag tree?
> 
> This was my fault for taking a shortcut.  I should have changed the name
> so the old add_taint worked still (set_taint?), then remove add_taint
> after the merge.
> 
> But I won't be rebasing, so you should be fine to merge it.

No worries.

Thanks
James

Re: linux-next build conflict between modules and metag trees (LOCKDEP_NOW_UNRELIABLE)

[Vineet Gupta]

On Thursday 07 February 2013 04:46 PM, James Hogan wrote:
> Hi Rusty,
> 
> The metag architecture tree adds an add_taint(TAINT_DIE) like other
> architectures do, and the modules-next tree adds the
> LOCKDEP_NOW_UNRELIABLE flag to all uses of add_taint (but obviously
> misses arch/metag since it doesn't exist yet), causing a compile error
> on metag in -next when the two are merged together.
> 
> Is it okay for me to merge your commit 373d4d0 ("taint: add explicit
> flag to show whether lock dep is still OK.") in modules-next into the
> base of the metag tree and expect it not to be rebased, so that I can
> then squash the fix into the metag tree?
> 
> The only commits this would include are:
> $ git log --oneline linus/master..373d4d0
> 373d4d0 taint: add explicit flag to show whether lock dep is still OK.
> 64748a2 module: printk message when module signature fail taints kernel.
> 
> Thanks
> James
> 

Being in the same situation as metag (ARC Port), what's the recommended practice
here - do we simply cherry-pick these changes into our tree - or do we merge the
"other" tree on top - ofcourse with premise that "other" tree will not rebase.

Thx,
-Vineet

Re: linux-next build conflict between modules and metag trees (LOCKDEP_NOW_UNRELIABLE)

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 1639 bytes --]

Hi Vineet,

On Fri, 8 Feb 2013 10:46:24 +0530 Vineet Gupta <Vineet.Gupta1@synopsys.com> wrote:
>
> On Thursday 07 February 2013 04:46 PM, James Hogan wrote:
> > 
> > The metag architecture tree adds an add_taint(TAINT_DIE) like other
> > architectures do, and the modules-next tree adds the
> > LOCKDEP_NOW_UNRELIABLE flag to all uses of add_taint (but obviously
> > misses arch/metag since it doesn't exist yet), causing a compile error
> > on metag in -next when the two are merged together.
> > 
> > Is it okay for me to merge your commit 373d4d0 ("taint: add explicit
> > flag to show whether lock dep is still OK.") in modules-next into the
> > base of the metag tree and expect it not to be rebased, so that I can
> > then squash the fix into the metag tree?
> > 
> > The only commits this would include are:
> > $ git log --oneline linus/master..373d4d0
> > 373d4d0 taint: add explicit flag to show whether lock dep is still OK.
> > 64748a2 module: printk message when module signature fail taints kernel.
> 
> Being in the same situation as metag (ARC Port), what's the recommended practice
> here - do we simply cherry-pick these changes into our tree - or do we merge the
> "other" tree on top - ofcourse with premise that "other" tree will not rebase.

Merging is better, as then the commits only exist once when your tree
gets merged back into Linus' tree.  However, such a merge should explain
why it is being done.  Assuming that the thing you merge does not get
rebased - which in this case, Rusty has said it won;t be.

-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au

[-- Attachment #2: Type: application/pgp-signature, Size: 836 bytes --]